Inaccurate measures of poverty: The problem with reporting ‘poverty rate drop’ without referencing the metric

A recently published column, courtesy of a certain PostMedia News  pundit, criticized the Canadian media for not reporting a June 27, 2013 Statscan release.

“’Poverty rate’ drops to lowest level ever and media turns a blind eye” read the headline on Andrew Coyne’s piece on July 22nd.

Briefly, there are a number of assumptions and methodologies underlying Statscan’s low-income measures, including  the LICO-AT (after-tax low-income cut-offs), that suggest they shouldn’t be used as measures of poverty. In case it wasn’t obvious to some, Statscan specifically cautions readers that the LICOs are not measures of poverty. The author seems to be aware of this, conveniently placing quotes around the term ‘poverty rate’ in the title.

Had the author consulted a few knowledgeable business/ economy journos, he likely would have had the answer to the seeming media silence. In summary:

1. The media is waiting for the 2011 NHS Income and Housing Release, scheduled for August 14, 2013.

2. The media is increasingly skeptical/critical of the Statscan releases, specifically the validity of both the data and metrics used.

Long-form Census and low-income

The Survey of Consumer Finances (SCF) and, more recently, Survey of Labour and Income Dynamics (SLID) were (past-tense; check the links for details) used to estimate income for the purpose of reporting annual low-income incidence. The Survey of Family Expenditures (Famex) and, more recently, Survey of Household Spending (SHS)1, have been used as the basis for the annually updated low-income cut-offs.

The one thing all those surveys had in common was they were all voluntary. And that was OK, so long as there were other surveys that could be relied on to (re)weigh the voluntary response samples to make them representative of the general population. The most important among those were the short and long-form Census, the only mandatory household surveys in Canada according to the Statistics Act. Or so most people assumed (see s.8 and s.19 of the Act), until the long-form Census was cancelled and replaced by the voluntary 2011 National Household Survey in 2010.

The final release of the Census cycle, Income and earnings (along with Housing and shelter costs), was always the most anticipated by the media. The long-form Census provided the most reliable and comprehensive measure of income, broken down by individual and household demographics, publicly available in Canada. The big headline grabbers were the direction and magnitude of income changes, along with changes in low-income incidence, relative to prior Census year(s).

Interestingly, the same ‘news’ outlet chastising the media for not reporting a relatively inconsequential July 27 Statscan release was the only one to take exception to Statscan’s report on stagnating middle-class incomes in its 2006 Census Income release.  ‘StatsCan sets off its own class war,’2 read the headlines of a Terence Corcoran column on May 2nd, 2008.

But back to the subject of the long-form Census and low income. As noted, it has been the biggest headline of the biggest release of the Census cycle (see 199620012006 Census Income releases).  Among the notable headlines the 2006 release generated were: “Almost 900,000 children live in poverty” from Loryanne Anthony for The Canadian Press; “The Canadian dream? 25 YEARS, 53 BUCKS” from Michael Valpy in The Globe and Mail.

Media anticipation of 2011 NHS Income release

Most of the detailed low income research in Canada, including Statscan’s own, references the Census low-income data for the reason already mentioned:it was the most comprehensive available.

Most experienced business / economy writers know this, which likely explains the scant coverage given  the June 27, 2013 low-income release from the Survey of Labour and Income Dynamics: the media is waiting for the major 2011 NHS low-income release scheduled for August 14, 2013.

If the last two 2011 NHS releases are indicative, the coverage will likely focus less on the Statscan’t analysis and more on how suspect the data is.  Readers may recall that the 2011 NHS immigration data was not in line with administrative data from the Department of Citizenship and Immigration Canada. As goes immigration, so goes visible minority data. And more than a few people questioned the 2011 NHS Aboriginal population estimates; As John Kolkman succinctly put it, This makes no sense.

And, one suspects, neither will the low income data from the 2011 NHS. Recent immigrants, visible minorities and Aboriginals tend to be among the disadvantaged groups over-represented (relative to their population share) in the low-income data. That these groups are no longer accurately represented in the population data will make the low-income data less reliable.

Media scepticism

Following The Great Depression, Canadians have seen spikes in household debt, precarious employment, lay-offs  – which newsrooms weren’t spared,  and social assistance at the same time government income security programs were clawed back in the name of austerity. The anecdotal evidence would make any reasonable person suspect of data indicating Canadian low-income incidence fell significantly over the same period.

Scepticism in the media is a good thing; It encourages more in-depth research and analysis. Unlike our seemingly oblivious Postmedia pundit, most business / economy journos likely looked up the 2011 LICO After-tax table and realised that the cut-offs were completely meaningless as a measure of poverty. As mentioned, no reasonable person would conclude that a couple living in a major Canadian urban centre could get by on $23K annually.

Some probably dug a little deeper and discovered that Statscan has increasingly adopted measures of low income that seem to serve no purpose other than to reduce the reported low income incidence (see here). All the more disappointing, given that the U.S. Census Bureau has taken the opposite approach with its Supplemental Poverty Measure (see Figure 7 in Income, Poverty, and Health Insurance Coverage in the United States: 2011 – PDF).


In closing, a note on data ‘cherry-picking’: Our perturbed PostMedia pundit has chastised the media for not reporting a data point which he asserts is indicative of a decline in Canadian poverty. However, he does not hold the same view of data and data sources that do not reconcile with his perspective. In fact, when several long-standing federal government advisories that compiled and published statistics on actual poverty were axed, he supported the decision.

“’Government isn’t obliged to bankroll its critics” read the headline on Andrew Coyne’s piece on May 4th, 2012.

Intellectually dishonest perhaps best sums up this approach to reporting. Then again, the PostMedia News pieces referenced in the preceding were all conveniently written as Op-Eds, not news items.


1 The low-income cut-offs are relative to average family expenditure based on the Survey of Household Spending. The survey’s methodology appears to excludes wealthier households (those that spend 70% or less of their earnings in a given year), likely biasing the cut-offs downward (see here).

2 In addition to more women working in 2005 relative to 1980, family composition, the proportion of the population living in a family, not to mention Statscan’s definition of family, all markedly changed over the referenced 25 year period. That’s why individual income is the preferred unit of historical comparison, and why the criticised Statscan report used it instead of family income.

Leave a Reply

Your email address will not be published. Required fields are marked *